> However, a TLB miss is usually fixed up with just a few memory
> references (if the page tables are in memory). This shouldn't
> require complicated routines or esoteric data structures; why
> not just directly check the Linux page tables and load the TLB
> from there, and ignore the hash/inverted-page-table? I'll admit,
> I'm PPC-challenged.
On a 603, you can do this fairly easily and automatically, though there
are some aspects that might be sticky (I haven't studied the 603 mechanism
recently). on a 604, I believe you could set up the inverted-page-table
to always not match and always convert the TLB reload to page fault, I
haven't thought it through well enough to be sure, but even if it would
work it would probably be inefficient. I really need to keep a copy of
the 604 spec at home.
Ed Welbon; welbon@bga.com;
"He had bought a large map representing the sea,
Without the least vestige of land:
And the crew was much pleased when they found it to be
A map they could all understand."